Introduction
Every forecast starts with history. For most businesses, that history is sales: invoices raised, orders shipped, units billed. It is the cleanest data in the ERP, so it becomes the foundation of the demand plan.
Sales history records something specific, though. It records what the business was able to supply. When a product ran out mid-month, the sales line shows the units that shipped. The orders that went unfilled, moved to a competitor or rolled into next month sit somewhere else, or nowhere at all.
Picture a month where customers wanted 1,000 units, the warehouse shipped 500, and closing stock ended at 10. A forecast built on sales sees 500 units of demand. A forecast built on true demand sees 1,000 units and a supply gap of 500. Those two numbers lead to very different production, inventory and dispatch plans.
This guide explains where true demand hides in your data, how to rebuild it, and how to start planning with it even when your data lives in spreadsheets, emails and several systems at once.
What is true demand in supply chain planning?
True demand, also called unconstrained demand, is the quantity customers would have bought if stock, capacity and delivery were fully available. Sales history is constrained demand: it reflects true demand after supply limits have shaped it.
The difference matters because a forecast learns from whatever history it is given. Feed it constrained sales, and it learns the supply limits along with the customer behaviour.
Sales, orders and demand: three different numbers
Signal | What it captures | Where it usually lives |
|---|---|---|
Sales (invoiced or shipped) | Units the business delivered | ERP billing and dispatch records |
Orders | Units customers asked for, including amendments | Order management, email, distributor portals |
True demand | Units customers wanted, including unfilled and lost orders | Spread across orders, stock records, sales notes and market inputs. |
Why the gap appears
- Stockouts: when a SKU is out of stock, sales stop while demand continues.
- Allocation: when supply is short, stock goes to priority channels or customers first, so other channels look weaker than they are.
- Order changes: amendments, cancellations and re-orders within short lead times blur what the customer first asked for.
- Returns: high return volumes in some categories and seasons mean gross sales overstate what customers kept.
Where true demand hides in your data
True demand is rarely missing. It is usually spread across records that were never designed to be read together. Five places hold most of it.
1. Stock-out periods inside sales history
The clearest signal is a period where sales dropped while closing stock sat near zero. That pattern points to a supply limit. Flagging these periods, and estimating what sales would have been, is the single most useful correction a demand plan can make.
2. Order amendments and short lead times
In e-commerce and modern trade, orders change often and lead times can be a week or less. The first order a customer places is the closest record of their intent. Later amendments often reflect what the supplier could confirm, so keeping the original order alongside the final one preserves the demand signal.
3. Returns and reverse flows
Some categories see return volumes rise with season or weather. Forecasting gross sales and returns separately gives a truer picture of net demand and helps size reverse logistics and storage.
4. Quarter-end and month-end peaks
Many businesses see a large share of volume land in the last weeks of a quarter. These peaks reflect sales targets and billing cycles as much as customer consumption. Smoothing them, or modelling them as a known pattern, stops the forecast from treating a billing rhythm as market demand.
5. Data split across systems
The rest of the signal sits in places a forecasting tool rarely sees:
- Spreadsheets maintained by sales and planning teams
- Email and WhatsApp threads with distributors and transporters
- Separate records for sales and after-sales service that do not link to each other
- Supplier and co-packer updates shared as files rather than data feeds
How to build a true demand signal, step by step
Rebuilding true demand is a sequence of corrections applied to sales history before any forecasting model runs. Each step is simple on its own; the value comes from applying them together and keeping them running every cycle.
- Bring sales, stock and orders into one view. Line up sales, closing stock and orders by SKU, location and week. This one table makes most supply-constrained periods visible.
- Flag constrained periods automatically. Mark any period where stock fell below a set threshold, or where orders exceeded shipments. These are the periods where sales understate demand.
- Estimate the missing demand. For flagged periods, estimate what sales would have been using the SKU's run rate before the stockout, the trend of similar SKUs, or sales at nearby locations that had stock.
- Separate one-off events from patterns. Tag promotions, bulk orders, price changes and quarter-end pushes so the model learns them as events rather than as baseline demand.
- Add outside signals. Seasonality, weather, festivals and regional events explain a large part of demand movement in consumer, agri and industrial categories. Layering them onto cleaned history sharpens the forecast.
- Capture planner context as data. When the sales team knows a distributor stocked up early or a competitor ran out, record it as a structured remark against the SKU and period. Over time, these remarks become inputs the model can use.
- Measure accuracy against true demand. Track forecast accuracy against the rebuilt demand figure as well as against sales, so the team can see whether the forecast is learning the market or the supply limits.
Planning with imperfect data: start where your data is today
You can start true demand forecasting with the data you have now. Waiting for a perfect ERP, a full data lake or complete master data delays the gains a better forecast brings, and the cleanup itself goes faster once a planning process is using the data every week.
What "good enough" data looks like
Data | Minimum to start | Improves over time with |
|---|---|---|
Sales history | 18 to 24 months by SKU and location, from ERP or Excel exports | Weekly granularity, channel split |
Stock records | Month-end or week-end closing stock | Daily stock, stock in transit |
Orders | Order quantity and order date | Original order kept alongside amendments |
Product master | SKU codes and product family | Lifecycle status, substitutes, root part numbers |
External signals | Holiday and festival calendar | Weather, regional events, market indices |
Practical ways to work with fragmented data
- Accept file uploads as a first step. Excel and CSV exports from SAP, Oracle or smaller ERPs are a valid starting point. Live API integrations can follow once the process proves its value.
- Clean as part of planning. Duplicate SKUs, unit mismatches and missing periods surface naturally once data feeds a weekly forecast, which makes them easier to fix than in a one-time cleanup project.
- Link records that belong together. Connecting sales records to service, warranty or returns data, even partially, reveals demand patterns that neither dataset shows alone.
- Use conversational AI to reach scattered data. An AI layer that reads ERP data, spreadsheets and email can answer "what was our stock position when this SKU's sales dropped?" in seconds, without a manual search across systems.
Turning true demand into better supply chain decisions
A true demand forecast pays off when it reaches the decisions downstream of it. Three areas benefit most.
Inventory planning and safety stock
Safety stock sized on constrained sales stays low for the very SKUs that ran out most often, which keeps the cycle going. Sizing it on true demand places buffers where customers actually need them, and frees stock from slow movers that only looked steady because their history was never interrupted.
Production planning and scheduling
When the demand plan separates customer demand from supply gaps, production teams can see which shortfalls came from capacity, which came from raw material, and which came from the plan itself. That clarity supports a more realistic master production schedule and fewer last-minute changes for export and priority orders.
Dispatch and replenishment
True demand by location shows where stock should sit before orders arrive. Replenishment and dispatch plans built on it move product to the depots and channels that will sell it, which reduces cross-warehouse transfers and urgent shipments.
From forecast to action with AI agents
Agentic AI is moving demand planning from a monthly report to a daily conversation. An AI agent connected to ERP, stock and order data can flag a SKU where sales dropped while stock hit zero, estimate the lost demand, and suggest a replenishment or production change for approval. The planning team stays in control of the decision, with the analysis done before they ask.
Conclusion
Sales history tells you what your supply chain delivered. True demand tells you what your market asked for. Building the second from the first takes a clear method, a few connected data sources and a planning cycle that keeps learning.
The businesses that make this shift plan inventory where customers need it, schedule production with fewer surprises and dispatch stock to the locations that will sell it. And they can start with the spreadsheets and ERP exports they already have.
See what your true demand looks like. Book a demo with Enmovil and explore your own sales, stock and order data in one planning view.
Frequently Asked Questions
What is the difference between sales and demand in forecasting?
What is unconstrained demand forecasting?
How do you forecast demand when stockouts distort sales history?
Can you do demand planning without an ERP or clean data?
How does AI improve demand forecasting accuracy?
What is demand sensing?
Share this article



